Research On Visual Inspection For Contour Defects Of Irregular Stamping Parts

With the upgrading of the manufacturing industry,irregular stamping parts have been widely used in aerospace,automobile,shipbuilding and other industrial areas.But its contour defects will not only affect the appearance of the product,but will even reduce the service life,reliability and safety of the final product.In order to avoid catastrophic accidents,the quality inspection of irregular stamping parts has become a hot and difficult issue in the manufacturing industry.At present,visual inspection has become an effective detection method with its advantages of objectiveness,high efficiency,accuracy,reliability,etc.It is of great value and significance to ensure product quality and improve the level of intelligent manufacturing in the stamping industry.To solve the problem of irregular stampings contour defects(irregular contour defects)detection,a complete inspection system based on machine vision is designed and three new visual detection methods for irregular contour defects are also put forward in this paper.The specific research contents are as follows:(1)According to the function and performance requirements of the contour defects detection system of irregular stamping parts,the hardware and software for the system are designed,and the overall scheme of the visual inspection system with irregular contour defects is introduced emphatically,including the selection and installation of hardware and the design of human-computer interface,so as to complete visual detection of contour defects more flexibly.(2)A visual detection algorithm for irregular contour defects based on Hu invariant moment is proposed.In view of the problem that the traditional contour detection algorithm can't combine anti-noise performance and positioning accuracy,the way of Chan-Vese level set,which does not rely on the image gradient,is carried out to detect the irregular contour of stamping parts.Then the Hu invariant moment of the obtained contour with rotation,translation and scaling invariance is calculated,and the matching rate of the two contours will be evaluated by the Hu invariant moment,to judge whether there is a contour defect in the part to be inspected.In order to improve the accuracy of Hu invariant moment contour matching,we adopt the way of multi-template matching.The experimental results show that the proposed method in this paper has a good detection effect.It can greatlyimprove the detection accuracy and meet the requirement of automatic detection for irregular contour defects of stamping parts in industry.(3)A Hough transform-based visual detection algorithm for irregular contour defects is put forward,mainly to solve two problems,such as: a)the image itself exists ambiguity due to pixels of target's edge and background with gradual transition;b)irregular stamping parts have singular shape,and the contour defects are small,which are difficult to detect.This method uses fuzzy set contour detection method to extract irregular contour,and image registration will be realized through Hough transform based on point and line feature of the contour.In this way,the contour defects can be identified,positioned and labeled effectively,through the similarity matching method,and the classification of irregular stamping parts can be also finished.(4)A method of image alignment-based matching for irregular contour defects visual inspection is presented.Aiming at the problem of high mismatch rate of traditional SIFT feature matching algorithm,a feature matching method based on cascade estimator is designed.It can effectively eliminate mismatched point pairs and estimate more accurate affine transformation parameters to realize image registration.Thus,the contour defects can be extracted simply by image difference and threshold segmentation,and the morphological technique is introduced to remove the interference of some pseudo-defects.Experimental results show that the proposed method can accurately identify the contour defects of irregular stamping parts and generate an accurate pixel-level defect location result of contour defects,which can mark out the contour defects.